A new study conducted by UCLA has proved that AI is more intelligent than humans. How? Researchers found that the autoregressive language model 'Generative Pre-trained Transformer-3 (GPT-3)' is the most compatible program concerning human intelligence.
The researchers at UCLA tested it out by running tests to compare its knowledge with an average college student. The program cleared all the tests with flying colors. It is even able to produce human-like text by incorporating deep learning. What astonishes us is that the program has 175 billion parameters, making it the largest program of its kind.
Now, how did the program compete with human intelligence? It was tested using three fronts. The first was SAT scores, followed by general knowledge, and lastly IQ. While humans only managed to achieve in the 35th percentile in the formerly mentioned fronts, the program managed to secure in the 80th percentile in all its tests.
GPT-3 performs complex tasks that aren’t that easy for AI, including abstract pattern induction. Most of the sites use such abstract patterns to rule out bots but with this development, now sites will have to come up with better ideas to keep AI out of their territories.
While this might have its disadvantages, the pros of having such a system overweigh the cons. Not only can this program be used for the development of better technology, but it can also contribute to the metaverse.
Something researchers are slightly concerned about is the process of acquiring this human-like knowledge. GPT-3 is a program with forced intelligence fed through mechanisms instead of training the system to learn it. This might cause it to either become too rigid or lead to confusion.
Experts also found that when solving problems from scratch and when solving MCQs, GPT-3 certainly outshined. This means the chances of error are comparatively lower which might give us a sigh of relief.
The only problem with developing such complex and advanced AI systems is the ‘robots taking over the World’ ideology. Although the theory does seem highly unlikely, with such emerging technologies, we do not know if it is so unattainable anymore.
GPT-3 is a huge accomplishment in the world of technology however we're still waiting for the study to be approved by fellow peers so that GPT-3 can finally make its grand entrance into the world.
Read next: Google Assistant, Siri, Alexa, And Bixby - The List For The World’s Best And Worst Voice Assistants Is Out
The researchers at UCLA tested it out by running tests to compare its knowledge with an average college student. The program cleared all the tests with flying colors. It is even able to produce human-like text by incorporating deep learning. What astonishes us is that the program has 175 billion parameters, making it the largest program of its kind.
Now, how did the program compete with human intelligence? It was tested using three fronts. The first was SAT scores, followed by general knowledge, and lastly IQ. While humans only managed to achieve in the 35th percentile in the formerly mentioned fronts, the program managed to secure in the 80th percentile in all its tests.
GPT-3 outperformed human participants on all problem types, both in terms of generative accuracy… and multiple-choice accuracy pic.twitter.com/KQge8cDQzh
— AI Breakfast (@AiBreakfast) December 27, 2022
GPT-3 performs complex tasks that aren’t that easy for AI, including abstract pattern induction. Most of the sites use such abstract patterns to rule out bots but with this development, now sites will have to come up with better ideas to keep AI out of their territories.
While this might have its disadvantages, the pros of having such a system overweigh the cons. Not only can this program be used for the development of better technology, but it can also contribute to the metaverse.
Something researchers are slightly concerned about is the process of acquiring this human-like knowledge. GPT-3 is a program with forced intelligence fed through mechanisms instead of training the system to learn it. This might cause it to either become too rigid or lead to confusion.
Experts also found that when solving problems from scratch and when solving MCQs, GPT-3 certainly outshined. This means the chances of error are comparatively lower which might give us a sigh of relief.
The only problem with developing such complex and advanced AI systems is the ‘robots taking over the World’ ideology. Although the theory does seem highly unlikely, with such emerging technologies, we do not know if it is so unattainable anymore.
GPT-3 is a huge accomplishment in the world of technology however we're still waiting for the study to be approved by fellow peers so that GPT-3 can finally make its grand entrance into the world.
Read next: Google Assistant, Siri, Alexa, And Bixby - The List For The World’s Best And Worst Voice Assistants Is Out
